DUPLICATE ONLINE PROPERTY LISTINGS IN RESIDENTIAL PROPERTY PRICE INDEX COMPILATION: A METHODOLOGICAL FRAMEWORK FOR UZBEKISTAN

Online real-estate listings provide timely and granular information for residential property price index compilation, but repeated advertisements for the same dwelling can inflate the number of observations, alter sample composition, and give some properties disproportionate influence in hedonic estimation. Methods: The study develops a property-quarter deduplication framework for Uzbekistan’s RPPI pipeline. It combines data standardization, blocking, similarity-based record linkage, gold-standard validation with Precision, Recall and F1, and parallel RAW-versus-DEDUP index calculations. The approach is designed for the existing strata-based, four-quarter rolling time-dummy hedonic system. Results: The empirical compilation context covers 1,654,058 listings for 2024Q1-2026Q2, 28 strata, seven rolling windows and 196 hedonic regressions. A duplicate-rate measure and a Duplicate Impact Diagnostic are specified, together with rules for keeping the strata structure, regression specification, outlier treatment, chaining and aggregation weights identical across parallel calculations. Because the current database does not retain duplicate labels or matched RAW and DEDUP samples, the actual index effect of deduplication is not reported as an estimated result. Conclusion: Deduplication should be treated as a statistical quality-control component rather than a simple technical cleaning step. A validated property-quarter record-linkage module would allow official compilers to document matching quality and quantify whether repeated listings materially affect RPPI dynamics.